Privacy in Identity & Access Management Systems

Privacy in Identity & Access Management systems Andreas Pashalidis Katholieke Universiteit Leuven, Belgium Chris J. Mitchell Royal Holloway, University of London, UK ABSTRACT This chapter surveys the approaches for addressing privacy in open identity and access management systems that have been taken by a number of current systems. The chapter begins by listing important privacy requirements and discusses how three systems that are being incrementally deployed in the Internet, namely SAML 2.0, CardSpace, and eID, address these requirements. Subsequently, the findings of recent European research projects in the area of privacy for I&AM systems are discussed. Finally, the approach taken to address the identified privacy requirements by ongoing projects is described at a high level. The overall goal of this chapter is to provide the reader with an overview of the diversity of privacy issues and techniques in the context of I&AM. INTRODUCTION Identity and Access Management (I&AM) systems support access control, namely ensuring that access to certain resources is granted only if the requestor is properly authorized. For example, a company employee that accesses a company VPN (Virtual Private Network) while working from abroad is likely to be granted access by an access control system. Although I&AM systems are closely integrated with access control systems, their main function is to support the system administrators and the end users in performing maintenance procedures, such as managing access credentials, user roles, access rights, rights delegation, auditing, and relationships between organizational units, throughout the lifetime of the system. Over the last fifty years, many I&AM systems with a wide range of functions have been developed. Such systems are typically composed of a number of modules, each with a specific task. Some I&AM systems are as simple as a database with authorized username/password pairs, while others are complex distributed systems that could include sophisticated policy decision points, interconnection with business process engines, accounting and billing infrastructures, credential negotiation agents, customer relationship management systems, administrative interfaces for the lifetime management of comprehensive user profiles, and provisions for auditing. Many I&AM systems are closed, i.e. they are designed for environments where there is a single system provider, such as a company or government organization, that has a very strong relationship with the prospective users. 2 The focus of this chapter is open I&AM systems, i.e. systems that cover multiple organizations. In the context of such systems, users interact with a range of different organizations using one or more credentials. New users may be introduced into the system by multiple parties, or users may be able to independently create new accounts for themselves. In open systems there is clearly a need for interoperability, and thus standardization is probably more important than in closed systems; privacy also plays a central role. Users should, for example, be able to control the degree of dissemination of their personal information to organizations and other users. The particular focus of this chapter is the various degrees of privacy achieved by current open I&AM systems, and what issues need to be addressed in future such systems. PRIVACY REQUIREMENTS FOR I&AM SYSTEMS The need for user privacy in open I&AM system arises from the need to reduce the risks of unnecessary or otherwise unwanted disclosure of personal information. In recent years, legislation in Europe, both at EU and at national levels, has become an important driver for the introduction of privacy and transparency enhancing techniques within I&AM systems. This is because many of these laws require businesses to follow the principles of data minimization, data protection, and, in some cases, data retention. The data minimization principle requires that personal data is not disclosed to a transacting partner unless that information is strictly needed in order to carry out the transaction. In order to establish such strict necessity, the purpose of disclosure must be specified for each data item to be disclosed. Data protection and retention require that users have access to, and can update, their personal information when it is stored at an organization, but also that organizations have to keep records in a way that facilitates effective investigation of past transactions. In this context, „personal data‟ is any data that could potentially lead to the identification of an individual, even if this is only possible in combination with additional information. The following more concrete requirements arise from the requirement to minimize the personal data that is transferred between parties. We say that a privacy-preserving I&AM system should enable its users to: selectively disclose personal data to organizations and other users; create multiple identities or pseudonyms; attach different pieces of personal information to different identities; review data disclosed in the past; maintain different identities towards different organizations; formulate „sticky‟ policies that follow personal data and that govern under which conditions the data may be disclosed and used; minimize the amount of trust users are required to place in third parties and infrastructural components in general; and provide explicit consent for sharing personal information, and enable users to revoke previously given consent. Of course, achieving all the above in a usable manner, i.e. without placing too great a burden on users and system administrators, is very challenging. The above requirements can be roughly captured by the following criteria. They can be used to evaluate I&AM systems with respect to their privacy-friendliness. 3 Trust model Some I&AM systems are designed so that a remote entity, typically called the „Identity Provider‟ (IdP), stores and manages the user‟s personal information. Users are typically authenticated by an IdP, and are then able to access their own information and forward it to requesting parties. While this has the advantage of mobility – users may use the system from any computer and any location – this model also raises significant privacy issues. This is because the trusted party not only learns the personal data of the user, but will also gain information about the behaviour and relationships of the user, since other parties will refer to the trusted party every time they require user data and or assurances about user authenticity. Moreover, the trusted party must be relied upon not to assert that a user has been authenticated when this has not occurred, and/or to assert false information about the personal attributes of a user (as discussed in (Alrodhan & Mitchell, 2010)). Of course, the privacy issues arising from the use of a third party IdP can to some extent be mitigated if the user is able to choose which IdP to use. This issue of choice arose starkly in the case of Passport, Microsoft‟s initial attempt to solve the identity management problem by making itself the IdP for everyone. As has been widely documented, the notion of trusting Microsoft with large quantities of personal data gave rise to a widespread and violent negative reaction, which clearly took Microsoft by surprise (Kormann & Rubin, 2000). Indeed, this informed Microsoft‟s subsequent effort in this space, the CardSpace system, discussed later in this chapter. Of course to be effective, choice requires a rich ecosystem of entities prepared to act as IdPs, and this ecosystem is still at an early stage of evolution. Moreover, even as and when such an ecosystem develops, not all users will be equipped with the means to decide which IdPs they can trust with mission-critical personal data. An approach in which personal data, such as attributes and certificates, are stored on the user‟s own computer and are then disclosed directly to the parties that require it, is likely to be more privacy-friendly. Such an approach, however, is less convenient, since mobility is no longer guaranteed and the users may have to perform a greater number of administrative tasks. It also increases the importance of security management of the platform on which the user information is stored. Ultimately it all comes down to trust. Users will have to make a trust decision with regard to the handling of personal data, either in terms of the use of a trusted third party (e.g. an IdP) or a personal platform. Sadly, recent history suggests that this is a highly problematic issue, since users are known to make poor trust decisions with regard to the handling of personal data, as the many issues identified with social networking sites have proved (Hogben, 2007). Multiple unlinkable identities The notion of a user identity is commonly defined as the set of personal information for that user (attributes, certificates, credentials, and other statements concerning the individual). The user may „compose‟ one or more personal identities by grouping relevant pieces of personal information. Identities do not need to be consistent; for example, one identity may include the user‟s real name and address, whereas a pseudonym (e.g. a nickname) and address (or no address at all) may be included in another. The grouping of data into identities makes it easier for a user to switch between contexts or roles. Note that the literature sometime uses the term „virtual identity‟ in order to refer to such a composed identity (see, e.g. (Aguiar, 2010)). In closed systems users are typically restricted to a single identity, whereas open systems typically do not have such a restriction. In particular, if the I&AM system interacts with multiple organizations that can each identify users, it may be desirable for users to be able to use different identities with different organizations. Moreover, in order to achieve data minimization, mechanisms should be provided to 4 prevent collaborating organizations from linking a given user‟s profile at one organization with the same user‟s profile at another.

Load more